The novel coronavirus is a new type of virus, and its transmission characteristics are different from the previous virus. Based on the SEIR transmission model, this paper redefines the latent state as close contacts state, introduces an asymptomatic infection state, and considers the influence of time on the state transition parameters in the model, proposing a new transmission model. The experimental results show that the fitting accuracy of the model has significantly improved. Compared with the traditional model, the fitting error was reduced by 8.3%-47.6%. Also, this study uses the US epidemic data as the training set to predict the development of the US epidemic, and the forecast results show that the US epidemic cannot be quickly controlled in a short time. However, the number of active cases will usher in a rapid decline after August 2021.